Electric driving head for drilling machine and construction method and identification method of stratum identification model

Abstract

The invention discloses an electric driving head for a drilling machine, and a construction method and an identification method of a stratum identification model, and belongs to the field of coal mine tunnel drilling machinery, the electric driving head comprises a mounting seat, a speed reducer is arranged on the mounting seat, a water feeder is arranged at the axial front end of the speed reducer, and a permanent magnet synchronous motor is arranged at the axial rear end of the speed reducer; the permanent magnet synchronous motor comprises a shell, and a front end cover and a rear end cover are arranged at the two ends of the shell. A permanent magnet synchronous motor body is arranged in the shell, and an output shaft of the permanent magnet synchronous motor body extends out of the front end cover and then is fixedly connected with the axial rear end of the speed reducer; a permanent magnet synchronous motor direct-driven speed reducer is adopted to replace a hydraulic pump station, a motor and hydraulic components, so that the use of hydraulic oil is reduced, and the environment is protected; the permanent magnet synchronous motor direct-drive speed reducer is high in transmission efficiency and more energy-saving; accurate control over the rotating speed and the torque and intelligent drilling are easier to achieve, and the technical problem that in the prior art, a traditional hydraulic drilling rig rotating power head is low in efficiency is solved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of coal mine tunnel drilling machines, and particularly relates to an electric drive head for a drilling machine, a construction method of a stratum identification model and an identification method. BACKGROUND

[0002] The conventional hydraulic drilling machine rotary power head is driven by hydraulic pressure, and needs to be matched with a pump station, a valve block, an oil tank and other hydraulic drive systems. The system is complex, the transmission efficiency is low, the energy consumption is large, the hydraulic oil consumption is large, and the use cost is high. Safe, efficient and green mining is the only way for sustainable development of the coal industry, and digitization, intelligentization and greenization are the development direction of coal mining. In terms of reducing the use of hydraulic oil and hydraulic components and improving transmission efficiency, electrification is the best solution, and the realization of electric drive of the drilling machine power head is the key to the realization of electric drive of the drilling machine,

[0003] Therefore, a pre-positioned water convenient electric drive power head for a drilling machine is urgently needed. The electric drilling machine power head adopts a permanent magnet synchronous motor as a power source, and the torque output is improved through a planetary reducer. The whole system simplifies the energy conversion link, greatly improves the transmission efficiency, and has simple structure and easy intelligent control. Compared with the hydraulic drilling machine, it does not need frequent maintenance, reduces the use of oil, and the motor can adaptively change the power according to the load. Therefore, the electric drive power head has obvious advantages in efficiency, convenience, reliability, multifunctionality, controllability and economy. SUMMARY

[0004] In view of the deficiencies in the prior art, the purpose of the present application is to provide an electric drive head for a drilling machine, a construction method of a stratum identification model and an identification method, to solve the technical problem of low efficiency of the conventional hydraulic drilling machine rotary power head in the prior art.

[0005] To solve the above technical problems, the present application adopts the following technical solutions:

[0006] An electric drive head for a drilling machine, comprising a mounting seat, a speed reducer is arranged on the mounting seat, a water feeder is arranged at the axial front end of the speed reducer, and a permanent magnet synchronous motor is arranged at the axial rear end of the speed reducer;

[0007] The permanent magnet synchronous motor comprises a shell, front and rear end covers are arranged at both ends of the shell, a permanent magnet synchronous motor body is arranged inside the shell, and an output shaft of the permanent magnet synchronous motor body extends out of the front end cover and is fixedly connected with the axial rear end of the speed reducer;

[0008] The permanent magnet synchronous motor is equipped with a water distributor; a first water inlet pipe and a first water outlet pipe are provided between the water distributor and the front end cover of the permanent magnet synchronous motor; a second water inlet pipe and a second water outlet pipe are provided between the water distributor and the rear end cover of the permanent magnet synchronous motor; and a third water inlet pipe and a third water outlet pipe are provided between the water distributor and the housing of the permanent magnet synchronous motor.

[0009] It also includes a controller electrically connected to the permanent magnet synchronous motor, the controller being used to identify the formation type and adjust the speed of the permanent magnet synchronous motor according to the formation type.

[0010] This invention also includes the following technical features:

[0011] The permanent magnet synchronous motor and the reducer, as well as the reducer and the water supply device, are connected by uniform splines.

[0012] Furthermore, a method for constructing a formation identification model, executed in the controller of the electric drive head of the drilling rig, includes the following steps:

[0013] Step 1: Real-time acquisition of the three-phase stationary coordinate system current signal I of the swimming pool synchronous motor. a (t), I b (t) and I c (t), and the Clark transform is used to convert the three-phase stationary coordinate system current signal into a two-phase stationary coordinate system current signal I. α and I β A;

[0014]

[0015] Step two, use the following formula to evaluate the two-phase stationary coordinate system current signal I obtained in step two. α and I β Normalization and preprocessing are performed to obtain the normalized signal. and A;

[0016]

[0017] in:

[0018] μ a μ β These are the two-phase stationary coordinate system current signals I. α and I β The mean, A;

[0019] σ a , σ β These are the two-phase stationary coordinate system current signals I. α and I β The standard deviation of A;

[0020] Step three, performing short-time Fourier transform on the normalized signal and and superimposing the frequency spectrum of the normalized signal and to obtain a two-dimensional feature matrix S;

[0021] Step four, constructing a formation identification model;

[0022] The formation identification model comprises an input layer, a convolution layer, a pooling layer, a full connection layer and an output layer connected in sequence;

[0023] The input layer is used for inputting the two-dimensional feature matrix S;

[0024] The convolution layer is used for extracting the local spatial relationship of the two-dimensional feature matrix S through convolution operation, capturing important time-frequency features, and outputting a feature map;

[0025] The pooling layer is used for down-sampling the feature map to obtain a smaller size feature map;

[0026] The full connection layer is used for feature unfolding of the smaller size feature map output by the pooling layer to obtain a plurality of one-dimensional vectors, integrating the plurality of one-dimensional vectors, and outputting a one-dimensional vector containing the original prediction score of each formation label;

[0027] The output layer is used for converting the one-dimensional vector containing the original prediction score of each formation label output by the full connection layer into a probability distribution through a Softmax function, so as to obtain the final prediction probability of the current input two-dimensional feature matrix S belonging to each formation label The maximum value of the final prediction probability corresponding to the formation label;

[0028]

[0029] Wherein:

[0030] yk is the original prediction score of the current input two-dimensional feature matrix S belonging to the kth formation label by the formation identification model;

[0031] e yk is an exponential operation on yk, which converts the corresponding original prediction score into a positive number;

[0032] yj is the original prediction score of the current input two-dimensional feature matrix S belonging to the jth formation label by the formation identification model;

[0033] e yj is an exponential operation on yj, which converts the corresponding original prediction score into a positive number;

[0034] K is the total number of stratum labels;

[0035] is the predicted probability of the kth stratum label;

[0036] Step five, using the two-dimensional feature matrix S obtained in step three as input and the corresponding stratum label as output, the stratum recognition model constructed in step four is trained using the cross-entropy loss function to obtain a trained stratum recognition model.

[0037] A stratum recognition method, executed in the controller of the electric drive head of the drilling rig, based on the construction method of the stratum recognition model, specifically comprising the following steps:

[0038] Step 1, by real-time acquisition of the three-phase stationary coordinate system current signals I a (t), I b (t) and I c (t) of the pool synchronous motor, and using clark transformation to convert the three-phase stationary coordinate system current signals into two-phase stationary coordinate system current signals I α and I β , A;

[0039]

[0040] Step 2, the two-phase stationary coordinate system current signals I α and I β obtained in step two are normalized and preprocessed using the following formula to obtain normalized signals and A;

[0041]

[0042] wherein:

[0043] μ a , μ β are the mean values of the two-phase stationary coordinate system current signals I α and I β , A;

[0044] σ a , σ β are the standard deviations of the two-phase stationary coordinate system current signals I α and I β , A;

[0045] Step 3, short-time Fourier transform is performed on the normalized signals and , and the frequency spectra of the normalized signals and are superimposed to obtain a two-dimensional feature matrix S;

[0046] Step 4, input the two-dimensional feature matrix S obtained in step 3 into the trained stratum identification model obtained in the method for constructing the stratum identification model, and output the corresponding stratum label;

[0047] Step 5, adjust the speed n of the permanent magnet synchronous motor according to the stratum label identified by the stratum identification model in step 4, and the adjustment amount is Δn.

[0048] Further comprising an adaptive control system for calculating the speed adjustment amount Δn of the permanent magnet synchronous motor, characterized by comprising an input layer, a hidden layer and an output layer arranged in sequence;

[0049] The input layer is used for inputting input variables, and the input variables include three-phase stationary coordinate system current signals and stratum labels;

[0050] The hidden layer adopts a nonlinear activation function ReLU and input variables to obtain output variables;

[0051] The output layer is used for outputting output variables, and the output variables include the speed adjustment amount Δn of the permanent magnet synchronous motor.

[0052] Compared with the prior art, the beneficial technical effects of the present application are:

[0053] (I) In the present application, the permanent magnet synchronous motor direct drive speed reducer replaces the hydraulic pump station, motor and hydraulic components, reduces the use of hydraulic oil, is more environmentally friendly, the permanent magnet synchronous motor direct drive speed reducer has high transmission efficiency, is more energy-saving, and is easier to realize accurate control of speed and torque and intelligent drilling, solving the technical problem of low efficiency of the rotary power head of the traditional hydraulic drilling rig in the prior art.

[0054] (II) In the present application, the change of the current signal of the permanent magnet synchronous motor accurately perceives the bottom stratum condition in the drilling process, and simultaneously realizes the corresponding strategy, increases the speed when encountering soft coal seams to improve the drilling efficiency, reduces the speed when encountering hard coal seams to prevent damage to the drill bit and drill pipe, and reduces the speed and retreats when encountering hard rock impact to protect the drilling tool.

[0055] (III) The adaptive controller in the present application can autonomously adjust the control strategy according to real-time data, adapt to different stratum conditions, has strong nonlinear mapping capability, can handle complex nonlinear relationships, and improves control accuracy. The controller has self-learning ability and can continuously optimize during operation. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 Fig. 1 is a three-dimensional structural schematic diagram of the water feeder electric drive power head of the drilling rig in the present application;

[0057] Figure 2Another angle three-dimensional structure schematic diagram of the electric drive head of the water feeder for the drilling rig in the application;

[0058] Figure 3 A two-dimensional structure schematic diagram of the electric drive head of the water feeder for the drilling rig in the application;

[0059] Figure 4 A control principle schematic diagram in the application;

[0060] Figure 5 A flow chart of the formation identification method in the application;

[0061] Fig. 6 is an effect schematic diagram in the verification example, wherein Fig. 6(a) is a neural network controller model training and verification loss function curve diagram, and Fig. 6(b) is a controller response difference diagram on the neural network and the traditional PID.

[0062] The meanings of various labels in the figure are as follows: 1 - mounting seat, 2 - speed reducer, 3 - water feeder, 4 - permanent magnet synchronous motor, 5 - water distributor, 6 - first water inlet pipe, 7 - first water outlet pipe, 8 - second water inlet pipe, 9 - second water outlet pipe, 10 - third water inlet pipe, 11 - third water outlet pipe;

[0063] 401 - shell, 402 - front end cover, 403 - rear end cover.

[0064] The specific content of the application is further explained and described in detail in combination with the embodiments below. DETAILED DESCRIPTION

[0065] It should be noted that all parts in the application, unless otherwise specified, use parts known in the art.

[0066] The specific embodiments of the application are given below, and it should be noted that the application is not limited to the following specific embodiments, and any equivalent variations made on the basis of the technical solutions of the application fall within the protection scope of the application.

[0067] The application provides an electric drive head for a drilling rig, which comprises a mounting seat 1, a speed reducer 2 is arranged on the mounting seat 1, a water feeder 3 is arranged at the axial front end of the speed reducer 2, and a permanent magnet synchronous motor 4 is arranged at the axial rear end of the speed reducer 2;

[0068] The permanent magnet synchronous motor 4 comprises a shell 401, front end covers 402 and rear end covers 403 are arranged at both ends of the shell 401; a permanent magnet synchronous motor body is arranged inside the shell 401, and an output shaft of the permanent magnet synchronous motor body extends out of the front end cover 402 and is fixedly connected with the axial rear end of the speed reducer 2;

[0069] The water distributor 5 is arranged on the permanent magnet synchronous motor 4; the first water inlet pipe 6 and the first water outlet pipe 7 are arranged between the water distributor and the front end cover 402 of the permanent magnet synchronous motor 4, the second water inlet pipe 8 and the second water outlet pipe 9 are arranged between the water distributor and the rear end cover 403 of the permanent magnet synchronous motor 4, and the third water inlet pipe 10 and the third water outlet pipe 11 are arranged between the water distributor and the shell 401 of the permanent magnet synchronous motor 4.

[0070] The controller is further included, and the controller is electrically connected with the permanent magnet synchronous motor 4, and the controller is used for identifying a formation category and adjusting a rotating speed of the permanent magnet synchronous motor through the formation category.

[0071] In the technical scheme, the permanent magnet synchronous motor directly drives the speed reducer to replace the hydraulic pump station, the motor and the hydraulic components, so that the use of hydraulic oil is reduced, and the environment is more friendly; the permanent magnet synchronous motor directly drives the speed reducer, so that the transmission efficiency is high, and energy is saved; the accurate control of the rotating speed and the torque and the intelligent drilling are more easily realized, and the technical problem of low efficiency of the rotary power head of the conventional hydraulic drilling rig is solved.

[0072] The water is arranged in front of the speed reducer, the torque of the motor is amplified through the speed reducer to realize the output of large torque, the water distributor 5 and the first water inlet pipe 6, the first water outlet pipe 7, the second water inlet pipe 8, the second water outlet pipe 9, the third water inlet pipe 10 and the third water outlet pipe 11 are arranged to realize the cooling of the permanent magnet synchronous motor 4, and the heat dissipation effect of the motor is ensured; the controller detects the current type of the permanent magnet synchronous motor to determine the formation information in the hole, and outputs a corresponding strategy to control the rotating speed of the permanent magnet synchronous motor, so that the drilling is smoothly implemented.

[0073] The permanent magnet synchronous motor 4 and the speed reducer 2 and the speed reducer 2 and the water feeder 3 are connected through the spline.

[0074] In the technical scheme, the spline connection is convenient for installation and disassembly.

[0075] A construction method of a formation identification model is executed in a controller of an electric drive head of a drilling rig, and includes the following steps:

[0076] Step one, the three-phase static coordinate system current signals I a (t), I b (t) and I c (t) of the permanent magnet synchronous motor are collected in real time, and the three-phase static coordinate system current signals are converted into two-phase static coordinate system current signals I α and I β through clark transformation;

[0077]

[0078] Step two, the two-phase stationary coordinate system current signal I α and I β are normalized and preprocessed to obtain normalized signals and A;

[0079]

[0080] wherein:

[0081] μ a , μ β are the mean values of the two-phase stationary coordinate system current signals I α and I β , A;

[0082] σ a , σ β are the standard deviations of the two-phase stationary coordinate system current signals I α and I β , A;

[0083] Step three, the normalized signals and are subjected to short-time Fourier transform, and the frequency spectra of the normalized signals and are superimposed to obtain a two-dimensional feature matrix S;

[0084] τ is a time variable in the short-time Fourier transform integral, representing the instantaneous time point corresponding to the window function sliding on the time axis within the entire signal duration;

[0085] Step four, a formation identification model is constructed;

[0086] The formation identification model comprises an input layer, a convolution layer, a pooling layer, a full connection layer and an output layer connected in sequence;

[0087] The input layer is used for inputting the two-dimensional feature matrix S;

[0088] The convolution layer is used for extracting the local spatial relationship of the two-dimensional feature matrix S through convolution operation, capturing important time-frequency features, and outputting a feature map;

[0089] The pooling layer is used for down-sampling the feature map to obtain a smaller size feature map;

[0090] The full connection layer is used for feature unfolding of the smaller size feature map output by the pooling layer to obtain a plurality of one-dimensional vectors, integrating the plurality of one-dimensional vectors, and outputting a one-dimensional vector containing the original prediction part of each formation label;

[0091] The output layer is used to convert the one-dimensional vector containing the original prediction scores of each formation label of the output of the full connection layer into a probability distribution through a Softmax function, so as to obtain the final prediction probability of the two-dimensional feature matrix S of the current input belonging to each formation label The maximum value of the final prediction probability Corresponds to the formation label.

[0092]

[0093] Wherein:

[0094] Yk is the original prediction score of the two-dimensional feature matrix S of the current input belonging to the kth formation label by the formation identification model;

[0095] e yk Exponential operation is performed on yk to convert the corresponding original prediction score into a positive number.

[0096] Yj is the original prediction score of the two-dimensional feature matrix S of the current input belonging to the jth formation label by the formation identification model;

[0097] e yj Exponential operation is performed on yj to convert the corresponding original prediction score into a positive number.

[0098] K is the total number of formation labels.

[0099] Yk is the prediction probability of the kth formation label.

[0100] Step five, using the two-dimensional feature matrix S obtained in step three as input and the corresponding formation label as output, the formation identification model constructed in step four is trained using a cross-entropy loss function to obtain a trained formation identification model.

[0101] In the above technical solution, the cross-entropy loss function is used to measure the difference between the model prediction result and the true formation label, and the gradient descent algorithm such as Adam optimizer is used to iteratively update the weights and bias parameters inside the model to gradually minimize the above difference until the model performance converges stably.

[0102] Generally, the size of the two-dimensional feature matrix S is selected as N*M to adapt to the input requirements of the CNN.

[0103] The pooling layer can reduce the data dimension, prevent overfitting, and retain significant features.

[0104] The present application provides a formation identification method, which is executed in the controller of the electric drive head of the drilling machine, and is based on the construction method of the formation identification model, and specifically includes the following steps:

[0105] Step 1, by real-time acquisition of three-phase stationary coordinate system current signals I a (t), I b (t) and I c (t), and adopting clark transformation to convert the three-phase stationary coordinate system current signals into two-phase stationary coordinate system current signals I α and I β , A;

[0106]

[0107] Step 2, the two-phase stationary coordinate system current signals I α and I β obtained in step two are normalized and preprocessed by adopting the following formula to obtain normalized signals and A;

[0108]

[0109] wherein:

[0110] μ a , μ β are the mean values of the two-phase stationary coordinate system current signals I α and I β , A;

[0111] σ a , σ β are the standard deviations of the two-phase stationary coordinate system current signals I α and I β , A;

[0112] Step 3, short-time Fourier transform is performed on the normalized signals and , and the frequency spectrums of the normalized signals and are superimposed to obtain a two-dimensional feature matrix S;

[0113] Step 4, the three-dimensional feature matrix S obtained in step 3 is input into the trained stratum identification model obtained by the stratum identification model construction method to output a corresponding stratum label;

[0114] Step 5, the speed n of the permanent magnet synchronous motor is adjusted according to the stratum label identified by the stratum identification model in step 4, and the adjustment amount is Δn.

[0115] Further comprising an adaptive control system for calculating the speed adjustment amount Δn of the permanent magnet synchronous motor, characterized by comprising an input layer, a hidden layer and an output layer arranged in sequence;

[0116] The input layer is used to input input variables, which include three-phase stationary coordinate system current signals and formation labels;

[0117] The hidden layer uses the non-linear activation function ReLU and the input variables to obtain the output variables;

[0118] The output layer is used to output output variables, including the speed adjustment Δn of the permanent magnet synchronous motor.

[0119] In the above technical solution, the adaptive controller can autonomously adjust the control strategy based on real-time data, adapt to different geological conditions, has strong nonlinear mapping capabilities, can handle complex nonlinear relationships, and improve control accuracy. The controller has self-learning capabilities and can continuously optimize during operation.

[0120] Example:

[0121] This embodiment presents a formation identification method, which involves real-time acquisition of the three-phase stationary coordinate system current signal I from a swimming pool synchronous motor. a (t), I b (t) and I c (t), and the Clark transform is used to convert the three-phase stationary coordinate system current signal into a two-phase stationary coordinate system current signal I. α and I β For the two-phase stationary coordinate system current signal I α and I β Normalization and preprocessing are performed to obtain the normalized signal. and

[0122] (Length 1024 points, sampling frequency 1024Hz, coverage 1s);

[0123] (Length 1024 points, and...) (Synchronous acquisition);

[0124] Then, using a window of length 128, the normalized signal is truncated by sliding with a 50% overlap rate. and Then perform Fourier transforms on each window segment, and finally take the amplitude value of the Fourier transform result to obtain the result. and and The short-time Fourier transform amplitude value S a and S β , will S a and S β The two-dimensional feature matrix S is obtained by superimposing time and frequency.

[0125] For example: The amplitude of the short-time Fourier transform of the 10th window, frequency point 15 (corresponding to f = 40 Hz) is 0.68, The amplitude of the short-time Fourier transform at the same position is 0.61, so S is S(10, 15) = 0.68 + 0.61 = 1.29, and the two-dimensional feature matrix S is obtained as follows:

[0126]

[0127] The above two-dimensional feature matrix S is input into the trained stratum identification model obtained by the stratum identification model construction method, and the corresponding stratum label output is a soft coal seam;

[0128] According to the stratum identification model, the stratum label obtained by the stratum identification model is adjusted to the speed n of the permanent magnet synchronous motor;

[0129] The current is 30A, the actual speed is 1200rmp, which is lower than the lower limit of the soft medium layer target, and the adjustment amount Δn = 10rmp according to the soft medium layer adjustment strategy; After adjustment, the motor current is improved, and the reacquisition is 33A, which meets the soft coal seam target current interval.

[0130] Verification example:

[0131] To verify the actual performance of the adaptive controller, simulation modeling and testing are carried out, and the following working conditions are set:

[0132] Initial speed n actual 1500rpm, three typical stratum categories: soft coal seam, hard coal seam, hard rock impact conditions, respectively test the response of the controller, the simulation results show that:

[0133] The control system realizes steady-state response within 0.5s, and the error convergence time is less than half of the traditional PID controller; The system overshoot is less than 5%, and there is no oscillation phenomenon, and the robustness is good; Under different stratum conditions, the attention mechanism makes the control strategy have obvious adaptability: the soft coal seam increases Δn, and the hard coal seam automatically limits the propulsion speed; The control error converges smoothly, and the dynamic error tends to zero, which verifies the effectiveness of real-time error feedback on network parameter adjustment.

[0134] To train the neural network controller model, a total of 3000 groups of historical speed, propulsion speed, current and stratum label data under different stratum conditions are collected as a training set. The training settings are as follows:

[0135] Network structure: 3 input nodes, 2 hidden layers (each with 16 neurons), 2 output nodes. Activation function: ReLU; Optimizer: Adam; Learning rate: 0.001, training rounds: 200 epochs, batch size: 64. The training results are as follows:

[0136] The loss function MSE decreased to below 0.002 within 100 epochs, and the final training accuracy was 98.2% and the verification accuracy was 96.7%.

[0137] In the hard rock impact sample, the error of Δn was less than ±3% on average, and the average error of the control response time and the actual set value was within ±2%. Figure 6(a) shows that the MSE decreases rapidly and tends to converge within 200 epochs, verifying that the model training effect is excellent; Figure 6(b) shows the difference in error response between the traditional PID and the neural network controller, and the neural network control response is smoother and converges faster.

Claims

1. An electric drive head for a drilling rig, comprising a mounting base (1), characterised in that, The mounting seat (1) is provided with a speed reducer (2), the axial front end of the speed reducer (2) is provided with a water feeder (3), and the axial rear end of the speed reducer (2) is provided with a permanent magnet synchronous motor (4); The permanent magnet synchronous motor (4) comprises a shell (401), the two ends of the shell (401) are provided with a front end cover (402) and a rear end cover (403), and the inside of the shell (401) is provided with a permanent magnet synchronous motor body; the output shaft of the permanent magnet synchronous motor body extends out of the front end cover (402) and is fixedly connected with the axial rear end of the speed reducer (2); The permanent magnet synchronous motor (4) is provided with a water distributor (5); the water distributor and the front end cover (402) of the permanent magnet synchronous motor (4) are provided with a first water inlet pipe (6) and a first water outlet pipe (7), the water distributor and the rear end cover (403) of the permanent magnet synchronous motor (4) are provided with a second water inlet pipe (8) and a second water outlet pipe (9), and the water distributor and the shell (401) of the permanent magnet synchronous motor (4) are provided with a third water inlet pipe (10) and a third water outlet pipe (11); A controller is further included, the controller is electrically connected with the permanent magnet synchronous motor (4), and the controller is used for identifying a formation category and adjusting the rotating speed of the permanent magnet synchronous motor through the formation category.

2. The electric drive head for a drilling rig of claim 1, wherein, The permanent magnet synchronous motor (4) and the speed reducer (2) and the speed reducer (2) and the water feeder (3) are connected by means of a spline.

3. A method of building a formation identification model, implemented in a controller of the electric drive head of the drilling rig of claim 1 or 2, characterized by, The method comprises the following steps: Step one, by real-time acquisition of three-phase stationary coordinate system current signal I a (t), I b (t) and I c (t), and using clark transformation to convert three-phase stationary coordinate system current signal into two-phase stationary coordinate system current signal I α and I β , A; Step two, use the following formula to evaluate the two-phase stationary coordinate system current signal I obtained in step two. α and I β Normalization and preprocessing are performed to obtain the normalized signal. and A; Wherein: μ a , μ β are the mean values of the two-phase stationary coordinate system current signals I α and I β , A; σ a , σ β are the standard deviations of the two-phase stationary coordinate system current signals I α and I β , respectively, A; Step three, performing a short-time Fourier transform on the normalized signal and stacking the spectra of the normalized signals and to obtain a two-dimensional feature matrix S; Step four, constructing a formation identification model; The formation identification model comprises an input layer, a convolution layer, a pooling layer, a full connection layer and an output layer connected in sequence; The input layer is used for inputting a two-dimensional feature matrix S; The convolution layer is used for extracting the local spatial relationship of the two-dimensional feature matrix S through a convolution operation, capturing important time-frequency features, and outputting a feature map; The pooling layer is used for downsampling the feature map to obtain a smaller size feature map; The full connection layer is used for feature expansion on the smaller size feature map output by the pooling layer to obtain a plurality of one-dimensional vectors, integrating the plurality of one-dimensional vectors, and outputting a one-dimensional vector containing the original prediction score of each formation label; The output layer is used to convert a one-dimensional vector containing individual stratum label original prediction scores of the output of the full connection layer into a probability distribution through a Softmax function, so as to obtain the final prediction probability of the two-dimensional feature matrix S of the current input belonging to each stratum label The output final prediction probability The stratum label corresponding to the maximum value of the final prediction probability Wherein: Yk is the original prediction score of the formation identification model for the current input two-dimensional feature matrix S belonging to the kth formation label; e yk To perform the exponentiation on yk, the corresponding original prediction score is converted to a positive number; Yj is the original prediction score of the formation identification model for the current input two-dimensional feature matrix S belonging to the jth formation label; e yj To perform the exponentiation on yj, the corresponding original prediction score is converted to a positive number; K is the total number of formation labels; is the predicted probability that the kth stratum label is correct; Step five, using the two-dimensional feature matrix S obtained in step three as input and the corresponding formation label as output, training the formation identification model constructed in step four by using a cross-entropy loss function to obtain a trained formation identification model.

4. A method for formation identification, which is executed in the controller of the electric drive head of the drilling rig according to claim 1 or 2, based on the method for constructing the formation identification model according to claim 3, characterized in that, Specifically comprising the following steps: Step one, by real-time acquisition of three-phase stationary coordinate system current signal I a (t), I b (t) and I c (t), and using clark transformation to convert three-phase stationary coordinate system current signal into two-phase stationary coordinate system current signal I α and I β , A; Step two, the two-phase stationary coordinate system current signal I obtained in step two is normalized and preprocessed by using the following formula α and I β to obtain normalized signals and A; Wherein: μ a , μ β are the mean values of the two-phase stationary coordinate system current signals I α and I β , A; σ a , σ β are the standard deviations of the two-phase stationary coordinate system current signals I α and I β , respectively, A; Step three, performing a short-time Fourier transform on the normalized signal and stacking the spectra of the normalized signals and to obtain a two-dimensional feature matrix S; Step 4, inputting the two-dimensional feature matrix S obtained in step 3 into the trained formation identification model obtained by the method for constructing the formation identification model in claim 3 to output the corresponding formation label; Step 5, adjusting the rotating speed n of the permanent magnet synchronous motor according to the formation label identified by the formation identification model in step 4, and the adjustment amount is Δn.

5. The method of claim 4, further comprising an adaptive control system for calculating a speed adjustment amount Δn of the permanent magnet synchronous motor, characterized in that, The input layer, the hidden layer and the output layer are sequentially arranged; The input layer is used for inputting input variables, and the input variables include three-phase stationary coordinate system current signals and formation labels; The hidden layer adopts a nonlinear activation function ReLU and input variables to obtain output variables; The output layer is used for outputting output variables, and the output variables include a speed adjustment amount Δn of the permanent magnet synchronous motor.

Images